1 /*
2 * XEN platform pci device, formerly known as the event channel device
3 *
4 * Copyright (c) 2003-2004 Intel Corp.
5 * Copyright (c) 2006 XenSource
6 *
7 * Permission is hereby granted, free of charge, to any person obtaining a copy
8 * of this software and associated documentation files (the "Software"), to deal
9 * in the Software without restriction, including without limitation the rights
10 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
11 * copies of the Software, and to permit persons to whom the Software is
12 * furnished to do so, subject to the following conditions:
13 *
14 * The above copyright notice and this permission notice shall be included in
15 * all copies or substantial portions of the Software.
16 *
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
21 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
22 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
23 * THE SOFTWARE.
24 */
25
26 #include "qemu/osdep.h"
27 #include "qapi/error.h"
28 #include "hw/ide/pci.h"
29 #include "hw/pci/pci.h"
30 #include "migration/vmstate.h"
31 #include "net/net.h"
32 #include "trace.h"
33 #include "sysemu/xen.h"
34 #include "sysemu/block-backend.h"
35 #include "qemu/error-report.h"
36 #include "qemu/module.h"
37 #include "qom/object.h"
38
39 #ifdef CONFIG_XEN
40 #include "hw/xen/xen_native.h"
41 #endif
42
43 /* The rule is that xen_native.h must come first */
44 #include "hw/xen/xen.h"
45
46 //#define DEBUG_PLATFORM
47
48 #ifdef DEBUG_PLATFORM
49 #define DPRINTF(fmt, ...) do { \
50 fprintf(stderr, "xen_platform: " fmt, ## __VA_ARGS__); \
51 } while (0)
52 #else
53 #define DPRINTF(fmt, ...) do { } while (0)
54 #endif
55
56 #define PFFLAG_ROM_LOCK 1 /* Sets whether ROM memory area is RW or RO */
57
58 struct PCIXenPlatformState {
59 /*< private >*/
60 PCIDevice parent_obj;
61 /*< public >*/
62
63 MemoryRegion fixed_io;
64 MemoryRegion bar;
65 MemoryRegion mmio_bar;
66 uint8_t flags; /* used only for version_id == 2 */
67 uint16_t driver_product_version;
68
69 /* Log from guest drivers */
70 char log_buffer[4096];
71 int log_buffer_off;
72 };
73
74 #define TYPE_XEN_PLATFORM "xen-platform"
OBJECT_DECLARE_SIMPLE_TYPE(PCIXenPlatformState,XEN_PLATFORM)75 OBJECT_DECLARE_SIMPLE_TYPE(PCIXenPlatformState, XEN_PLATFORM)
76
77 #define XEN_PLATFORM_IOPORT 0x10
78
79 /* Send bytes to syslog */
80 static void log_writeb(PCIXenPlatformState *s, char val)
81 {
82 if (val == '\n' || s->log_buffer_off == sizeof(s->log_buffer) - 1) {
83 /* Flush buffer */
84 s->log_buffer[s->log_buffer_off] = 0;
85 trace_xen_platform_log(s->log_buffer);
86 s->log_buffer_off = 0;
87 } else {
88 s->log_buffer[s->log_buffer_off++] = val;
89 }
90 }
91
92 /*
93 * Unplug device flags.
94 *
95 * The logic got a little confused at some point in the past but this is
96 * what they do now.
97 *
98 * bit 0: Unplug all IDE and SCSI disks.
99 * bit 1: Unplug all NICs.
100 * bit 2: Unplug IDE disks except primary master. This is overridden if
101 * bit 0 is also present in the mask.
102 * bit 3: Unplug all NVMe disks.
103 *
104 */
105 #define _UNPLUG_IDE_SCSI_DISKS 0
106 #define UNPLUG_IDE_SCSI_DISKS (1u << _UNPLUG_IDE_SCSI_DISKS)
107
108 #define _UNPLUG_ALL_NICS 1
109 #define UNPLUG_ALL_NICS (1u << _UNPLUG_ALL_NICS)
110
111 #define _UNPLUG_AUX_IDE_DISKS 2
112 #define UNPLUG_AUX_IDE_DISKS (1u << _UNPLUG_AUX_IDE_DISKS)
113
114 #define _UNPLUG_NVME_DISKS 3
115 #define UNPLUG_NVME_DISKS (1u << _UNPLUG_NVME_DISKS)
116
pci_device_is_passthrough(PCIDevice * d)117 static bool pci_device_is_passthrough(PCIDevice *d)
118 {
119 if (!strcmp(d->name, "xen-pci-passthrough")) {
120 return true;
121 }
122
123 if (xen_mode == XEN_EMULATE && !strcmp(d->name, "vfio-pci")) {
124 return true;
125 }
126
127 return false;
128 }
129
unplug_nic(PCIBus * b,PCIDevice * d,void * o)130 static void unplug_nic(PCIBus *b, PCIDevice *d, void *o)
131 {
132 /* We have to ignore passthrough devices */
133 if (pci_get_word(d->config + PCI_CLASS_DEVICE) ==
134 PCI_CLASS_NETWORK_ETHERNET
135 && !pci_device_is_passthrough(d)) {
136 object_unparent(OBJECT(d));
137 }
138 }
139
140 /* Remove the peer of the NIC device. Normally, this would be a tap device. */
del_nic_peer(NICState * nic,void * opaque)141 static void del_nic_peer(NICState *nic, void *opaque)
142 {
143 NetClientState *nc = qemu_get_queue(nic);
144 ObjectClass *klass = module_object_class_by_name(nc->model);
145
146 /* Only delete peers of PCI NICs that we're about to delete */
147 if (!klass || !object_class_dynamic_cast(klass, TYPE_PCI_DEVICE)) {
148 return;
149 }
150
151 if (nc->peer)
152 qemu_del_net_client(nc->peer);
153 }
154
pci_unplug_nics(PCIBus * bus)155 static void pci_unplug_nics(PCIBus *bus)
156 {
157 qemu_foreach_nic(del_nic_peer, NULL);
158 pci_for_each_device(bus, 0, unplug_nic, NULL);
159 }
160
161 /*
162 * The Xen HVM unplug protocol [1] specifies a mechanism to allow guests to
163 * request unplug of 'aux' disks (which is stated to mean all IDE disks,
164 * except the primary master).
165 *
166 * NOTE: The semantics of what happens if unplug of all disks and 'aux' disks
167 * is simultaneously requested is not clear. The implementation assumes
168 * that an 'all' request overrides an 'aux' request.
169 *
170 * [1] https://xenbits.xen.org/gitweb/?p=xen.git;a=blob;f=docs/misc/hvm-emulated-unplug.pandoc
171 */
172 struct ide_unplug_state {
173 bool aux;
174 int nr_unplugged;
175 };
176
ide_dev_unplug(DeviceState * dev,void * _st)177 static int ide_dev_unplug(DeviceState *dev, void *_st)
178 {
179 struct ide_unplug_state *st = _st;
180 IDEDevice *idedev;
181 IDEBus *idebus;
182 BlockBackend *blk;
183 int unit;
184
185 idedev = IDE_DEVICE(object_dynamic_cast(OBJECT(dev), "ide-hd"));
186 if (!idedev) {
187 return 0;
188 }
189
190 idebus = IDE_BUS(qdev_get_parent_bus(dev));
191
192 unit = (idedev == idebus->slave);
193 assert(unit || idedev == idebus->master);
194
195 if (st->aux && !unit && !strcmp(BUS(idebus)->name, "ide.0")) {
196 return 0;
197 }
198
199 blk = idebus->ifs[unit].blk;
200 if (blk) {
201 blk_drain(blk);
202 blk_flush(blk);
203
204 blk_detach_dev(blk, DEVICE(idedev));
205 idebus->ifs[unit].blk = NULL;
206 idedev->conf.blk = NULL;
207 monitor_remove_blk(blk);
208 blk_unref(blk);
209 }
210
211 object_unparent(OBJECT(dev));
212 st->nr_unplugged++;
213
214 return 0;
215 }
216
pci_xen_ide_unplug(PCIDevice * d,bool aux)217 static void pci_xen_ide_unplug(PCIDevice *d, bool aux)
218 {
219 struct ide_unplug_state st = { aux, 0 };
220 DeviceState *dev = DEVICE(d);
221
222 qdev_walk_children(dev, NULL, NULL, ide_dev_unplug, NULL, &st);
223 if (st.nr_unplugged) {
224 pci_device_reset(d);
225 }
226 }
227
unplug_disks(PCIBus * b,PCIDevice * d,void * opaque)228 static void unplug_disks(PCIBus *b, PCIDevice *d, void *opaque)
229 {
230 uint32_t flags = *(uint32_t *)opaque;
231 bool aux = (flags & UNPLUG_AUX_IDE_DISKS) &&
232 !(flags & UNPLUG_IDE_SCSI_DISKS);
233
234 /* We have to ignore passthrough devices */
235 if (pci_device_is_passthrough(d))
236 return;
237
238 switch (pci_get_word(d->config + PCI_CLASS_DEVICE)) {
239 case PCI_CLASS_STORAGE_IDE:
240 case PCI_CLASS_STORAGE_SATA:
241 pci_xen_ide_unplug(d, aux);
242 break;
243
244 case PCI_CLASS_STORAGE_SCSI:
245 if (!aux) {
246 object_unparent(OBJECT(d));
247 }
248 break;
249
250 case PCI_CLASS_STORAGE_EXPRESS:
251 if (flags & UNPLUG_NVME_DISKS) {
252 object_unparent(OBJECT(d));
253 }
254
255 default:
256 break;
257 }
258 }
259
pci_unplug_disks(PCIBus * bus,uint32_t flags)260 static void pci_unplug_disks(PCIBus *bus, uint32_t flags)
261 {
262 pci_for_each_device(bus, 0, unplug_disks, &flags);
263 }
264
platform_fixed_ioport_writew(void * opaque,uint32_t addr,uint32_t val)265 static void platform_fixed_ioport_writew(void *opaque, uint32_t addr, uint32_t val)
266 {
267 PCIXenPlatformState *s = opaque;
268
269 switch (addr) {
270 case 0: {
271 PCIDevice *pci_dev = PCI_DEVICE(s);
272 /* Unplug devices. See comment above flag definitions */
273 if (val & (UNPLUG_IDE_SCSI_DISKS | UNPLUG_AUX_IDE_DISKS |
274 UNPLUG_NVME_DISKS)) {
275 DPRINTF("unplug disks\n");
276 pci_unplug_disks(pci_get_bus(pci_dev), val);
277 }
278 if (val & UNPLUG_ALL_NICS) {
279 DPRINTF("unplug nics\n");
280 pci_unplug_nics(pci_get_bus(pci_dev));
281 }
282 break;
283 }
284 case 2:
285 switch (val) {
286 case 1:
287 DPRINTF("Citrix Windows PV drivers loaded in guest\n");
288 break;
289 case 0:
290 DPRINTF("Guest claimed to be running PV product 0?\n");
291 break;
292 default:
293 DPRINTF("Unknown PV product %d loaded in guest\n", val);
294 break;
295 }
296 s->driver_product_version = val;
297 break;
298 }
299 }
300
platform_fixed_ioport_writel(void * opaque,uint32_t addr,uint32_t val)301 static void platform_fixed_ioport_writel(void *opaque, uint32_t addr,
302 uint32_t val)
303 {
304 switch (addr) {
305 case 0:
306 /* PV driver version */
307 break;
308 }
309 }
310
platform_fixed_ioport_writeb(void * opaque,uint32_t addr,uint32_t val)311 static void platform_fixed_ioport_writeb(void *opaque, uint32_t addr, uint32_t val)
312 {
313 PCIXenPlatformState *s = opaque;
314
315 switch (addr) {
316 case 0: /* Platform flags */
317 if (xen_mode == XEN_EMULATE) {
318 /* XX: Use i440gx/q35 PAM setup to do this? */
319 s->flags = val & PFFLAG_ROM_LOCK;
320 #ifdef CONFIG_XEN
321 } else {
322 hvmmem_type_t mem_type = (val & PFFLAG_ROM_LOCK) ?
323 HVMMEM_ram_ro : HVMMEM_ram_rw;
324
325 if (xen_set_mem_type(xen_domid, mem_type, 0xc0, 0x40)) {
326 DPRINTF("unable to change ro/rw state of ROM memory area!\n");
327 } else {
328 s->flags = val & PFFLAG_ROM_LOCK;
329 DPRINTF("changed ro/rw state of ROM memory area. now is %s state.\n",
330 (mem_type == HVMMEM_ram_ro ? "ro" : "rw"));
331 }
332 #endif
333 }
334 break;
335
336 case 2:
337 log_writeb(s, val);
338 break;
339 }
340 }
341
platform_fixed_ioport_readw(void * opaque,uint32_t addr)342 static uint32_t platform_fixed_ioport_readw(void *opaque, uint32_t addr)
343 {
344 switch (addr) {
345 case 0:
346 /* Magic value so that you can identify the interface. */
347 return 0x49d2;
348 default:
349 return 0xffff;
350 }
351 }
352
platform_fixed_ioport_readb(void * opaque,uint32_t addr)353 static uint32_t platform_fixed_ioport_readb(void *opaque, uint32_t addr)
354 {
355 PCIXenPlatformState *s = opaque;
356
357 switch (addr) {
358 case 0:
359 /* Platform flags */
360 return s->flags;
361 case 2:
362 /* Version number */
363 return 1;
364 default:
365 return 0xff;
366 }
367 }
368
platform_fixed_ioport_reset(void * opaque)369 static void platform_fixed_ioport_reset(void *opaque)
370 {
371 PCIXenPlatformState *s = opaque;
372
373 platform_fixed_ioport_writeb(s, 0, 0);
374 }
375
platform_fixed_ioport_read(void * opaque,hwaddr addr,unsigned size)376 static uint64_t platform_fixed_ioport_read(void *opaque,
377 hwaddr addr,
378 unsigned size)
379 {
380 switch (size) {
381 case 1:
382 return platform_fixed_ioport_readb(opaque, addr);
383 case 2:
384 return platform_fixed_ioport_readw(opaque, addr);
385 default:
386 return -1;
387 }
388 }
389
platform_fixed_ioport_write(void * opaque,hwaddr addr,uint64_t val,unsigned size)390 static void platform_fixed_ioport_write(void *opaque, hwaddr addr,
391
392 uint64_t val, unsigned size)
393 {
394 switch (size) {
395 case 1:
396 platform_fixed_ioport_writeb(opaque, addr, val);
397 break;
398 case 2:
399 platform_fixed_ioport_writew(opaque, addr, val);
400 break;
401 case 4:
402 platform_fixed_ioport_writel(opaque, addr, val);
403 break;
404 }
405 }
406
407
408 static const MemoryRegionOps platform_fixed_io_ops = {
409 .read = platform_fixed_ioport_read,
410 .write = platform_fixed_ioport_write,
411 .valid = {
412 .unaligned = true,
413 },
414 .impl = {
415 .min_access_size = 1,
416 .max_access_size = 4,
417 .unaligned = true,
418 },
419 .endianness = DEVICE_LITTLE_ENDIAN,
420 };
421
platform_fixed_ioport_init(PCIXenPlatformState * s)422 static void platform_fixed_ioport_init(PCIXenPlatformState* s)
423 {
424 memory_region_init_io(&s->fixed_io, OBJECT(s), &platform_fixed_io_ops, s,
425 "xen-fixed", 16);
426 memory_region_add_subregion(get_system_io(), XEN_PLATFORM_IOPORT,
427 &s->fixed_io);
428 }
429
430 /* Xen Platform PCI Device */
431
xen_platform_ioport_readb(void * opaque,hwaddr addr,unsigned int size)432 static uint64_t xen_platform_ioport_readb(void *opaque, hwaddr addr,
433 unsigned int size)
434 {
435 if (addr == 0) {
436 return platform_fixed_ioport_readb(opaque, 0);
437 } else {
438 return ~0u;
439 }
440 }
441
xen_platform_ioport_writeb(void * opaque,hwaddr addr,uint64_t val,unsigned int size)442 static void xen_platform_ioport_writeb(void *opaque, hwaddr addr,
443 uint64_t val, unsigned int size)
444 {
445 PCIXenPlatformState *s = opaque;
446 PCIDevice *pci_dev = PCI_DEVICE(s);
447
448 switch (addr) {
449 case 0: /* Platform flags */
450 platform_fixed_ioport_writeb(opaque, 0, (uint32_t)val);
451 break;
452 case 4:
453 if (val == 1) {
454 /*
455 * SUSE unplug for Xenlinux
456 * xen-kmp used this since xen-3.0.4, instead the official protocol
457 * from xen-3.3+ It did an unconditional "outl(1, (ioaddr + 4));"
458 * Pre VMDP 1.7 used 4 and 8 depending on how VMDP was configured.
459 * If VMDP was to control both disk and LAN it would use 4.
460 * If it controlled just disk or just LAN, it would use 8 below.
461 */
462 pci_unplug_disks(pci_get_bus(pci_dev), UNPLUG_IDE_SCSI_DISKS);
463 pci_unplug_nics(pci_get_bus(pci_dev));
464 }
465 break;
466 case 8:
467 switch (val) {
468 case 1:
469 pci_unplug_disks(pci_get_bus(pci_dev), UNPLUG_IDE_SCSI_DISKS);
470 break;
471 case 2:
472 pci_unplug_nics(pci_get_bus(pci_dev));
473 break;
474 default:
475 log_writeb(s, (uint32_t)val);
476 break;
477 }
478 break;
479 default:
480 break;
481 }
482 }
483
484 static const MemoryRegionOps xen_pci_io_ops = {
485 .read = xen_platform_ioport_readb,
486 .write = xen_platform_ioport_writeb,
487 .impl.min_access_size = 1,
488 .impl.max_access_size = 1,
489 };
490
platform_ioport_bar_setup(PCIXenPlatformState * d)491 static void platform_ioport_bar_setup(PCIXenPlatformState *d)
492 {
493 memory_region_init_io(&d->bar, OBJECT(d), &xen_pci_io_ops, d,
494 "xen-pci", 0x100);
495 }
496
platform_mmio_read(void * opaque,hwaddr addr,unsigned size)497 static uint64_t platform_mmio_read(void *opaque, hwaddr addr,
498 unsigned size)
499 {
500 DPRINTF("Warning: attempted read from physical address "
501 "0x" HWADDR_FMT_plx " in xen platform mmio space\n", addr);
502
503 return 0;
504 }
505
platform_mmio_write(void * opaque,hwaddr addr,uint64_t val,unsigned size)506 static void platform_mmio_write(void *opaque, hwaddr addr,
507 uint64_t val, unsigned size)
508 {
509 DPRINTF("Warning: attempted write of 0x%"PRIx64" to physical "
510 "address 0x" HWADDR_FMT_plx " in xen platform mmio space\n",
511 val, addr);
512 }
513
514 static const MemoryRegionOps platform_mmio_handler = {
515 .read = &platform_mmio_read,
516 .write = &platform_mmio_write,
517 .endianness = DEVICE_NATIVE_ENDIAN,
518 };
519
platform_mmio_setup(PCIXenPlatformState * d)520 static void platform_mmio_setup(PCIXenPlatformState *d)
521 {
522 memory_region_init_io(&d->mmio_bar, OBJECT(d), &platform_mmio_handler, d,
523 "xen-mmio", 0x1000000);
524 }
525
xen_platform_post_load(void * opaque,int version_id)526 static int xen_platform_post_load(void *opaque, int version_id)
527 {
528 PCIXenPlatformState *s = opaque;
529
530 platform_fixed_ioport_writeb(s, 0, s->flags);
531
532 return 0;
533 }
534
535 static const VMStateDescription vmstate_xen_platform = {
536 .name = "platform",
537 .version_id = 4,
538 .minimum_version_id = 4,
539 .post_load = xen_platform_post_load,
540 .fields = (const VMStateField[]) {
541 VMSTATE_PCI_DEVICE(parent_obj, PCIXenPlatformState),
542 VMSTATE_UINT8(flags, PCIXenPlatformState),
543 VMSTATE_END_OF_LIST()
544 }
545 };
546
xen_platform_realize(PCIDevice * dev,Error ** errp)547 static void xen_platform_realize(PCIDevice *dev, Error **errp)
548 {
549 PCIXenPlatformState *d = XEN_PLATFORM(dev);
550 uint8_t *pci_conf;
551
552 /* Device will crash on reset if xen is not initialized */
553 if (xen_mode == XEN_DISABLED) {
554 error_setg(errp, "xen-platform device requires a Xen guest");
555 return;
556 }
557
558 pci_conf = dev->config;
559
560 pci_set_word(pci_conf + PCI_COMMAND, PCI_COMMAND_IO | PCI_COMMAND_MEMORY);
561
562 pci_config_set_prog_interface(pci_conf, 0);
563
564 pci_conf[PCI_INTERRUPT_PIN] = 1;
565
566 platform_ioport_bar_setup(d);
567 pci_register_bar(dev, 0, PCI_BASE_ADDRESS_SPACE_IO, &d->bar);
568
569 /* reserve 16MB mmio address for share memory*/
570 platform_mmio_setup(d);
571 pci_register_bar(dev, 1, PCI_BASE_ADDRESS_MEM_PREFETCH,
572 &d->mmio_bar);
573
574 platform_fixed_ioport_init(d);
575 }
576
platform_reset(DeviceState * dev)577 static void platform_reset(DeviceState *dev)
578 {
579 PCIXenPlatformState *s = XEN_PLATFORM(dev);
580
581 platform_fixed_ioport_reset(s);
582 }
583
xen_platform_class_init(ObjectClass * klass,void * data)584 static void xen_platform_class_init(ObjectClass *klass, void *data)
585 {
586 DeviceClass *dc = DEVICE_CLASS(klass);
587 PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
588
589 k->realize = xen_platform_realize;
590 k->vendor_id = PCI_VENDOR_ID_XEN;
591 k->device_id = PCI_DEVICE_ID_XEN_PLATFORM;
592 k->class_id = PCI_CLASS_OTHERS << 8 | 0x80;
593 k->subsystem_vendor_id = PCI_VENDOR_ID_XEN;
594 k->subsystem_id = PCI_DEVICE_ID_XEN_PLATFORM;
595 k->revision = 1;
596 set_bit(DEVICE_CATEGORY_MISC, dc->categories);
597 dc->desc = "XEN platform pci device";
598 device_class_set_legacy_reset(dc, platform_reset);
599 dc->vmsd = &vmstate_xen_platform;
600 }
601
602 static const TypeInfo xen_platform_info = {
603 .name = TYPE_XEN_PLATFORM,
604 .parent = TYPE_PCI_DEVICE,
605 .instance_size = sizeof(PCIXenPlatformState),
606 .class_init = xen_platform_class_init,
607 .interfaces = (InterfaceInfo[]) {
608 { INTERFACE_CONVENTIONAL_PCI_DEVICE },
609 { },
610 },
611 };
612
xen_platform_register_types(void)613 static void xen_platform_register_types(void)
614 {
615 type_register_static(&xen_platform_info);
616 }
617
618 type_init(xen_platform_register_types)
619